1. Field of the Invention
The present invention relates to a technology for distributing the load of a network using a cache technique.
2. Description of the Related Art
Recently, a cache technique that is effective when the content on a content server is not modified and a mirroring technique that is suitable when a large amount of data is copied in a certain cycle, among others, have been known as technologies for distributing the load of a network. These techniques reduce the amount of communication of content flowing on a network by providing a cache server on a network, copying information possessed by a content server to a cache server on a network close to a reference requester terminal (client) (for example, if it is a content distribution network), and enabling the client to refer to the copy on the cache server.
However, cases that cannot be addressed by simply caching content, such as a web server that dynamically generates content when a client accesses it, have increased when using the cache technique. As its solution, a technology for increasing the ratio of content that can be cached by elaborately dividing even content that is dynamically generated as a whole into a dynamic part and a static part and caching the parts other than dynamic one has been developed. There is also a technology for improving the hit rate of cache by caching dynamic content and also automatically modifying the cache when the content is modified.
However, since the load of a cache server increases according to the number of requests, a load-distributing technology is indispensable. Although there are methods using a load-distributing controller, using a wide-band load distribution and the like, they do not take into consideration a state where a cache server is managed by different managers and a state where files are consecutively copied to a cache and the load causes a problem.
In a peer-to-peer content delivery network (P2P-CDN) composed of a plurality of content servers, a cache server for receiving the distribution of content from the content servers and relaying the content and a client, and a technology for modifying the type of the content distributed by the content servers according to the change of the process load of the cache server (which is measured on the basis of the request of a client) is proposed.
However, since in the load distribution of content distribution in P2P-CDN and the like, the logical network configuration of a cache server and a client receiving content is generally fixed, the network configuration cannot be modified even when its load is unbalanced.
Patent Document 1 proposes a congestion control device for enabling the effective use of services comprising a plurality of client terminals capable of requesting information and displaying information obtained in response to it, a plurality of content servers for transmitting information in response to a request to a client terminal, and a congestion controller provided between them.
Patent document 2 proposes a system comprising a load measurement unit for measuring the load of a cache server, an overload determination unit for determining whether the load is overloaded by comparing the measured load with a predetermined value, a connecting destination retrieve request transmitting unit for transmitting connecting destination retrieve request information requesting a content server to search for the connecting destination of a load source cache server when the load is overloaded, a connecting destination information receiving unit for receiving the connecting destination information indicating the searched connecting destination from the transmitting source content server that has transmitted the connecting destination retrieve request information, and a switch request transmitting unit for transmitting switch request information requesting the load source cache server to switch the connection to the connection g destination indicated by the connecting destination information on the basis of the received connecting destination information wherein the amount of communications flowing through a network is reduced as a whole by distributing the load of each cache.
FIGS. 1A and 1B explain the problem of the prior art.
FIG. 1A explains a negotiation for equalizing the load of a network. CDN1 comprises a plurality of content servers 2, a plurality of cache servers 3, and a plurality of clients 4. The plurality of cache servers 3 is subordinated to each of the plurality of content servers 2 and the plurality of clients 4 is subordinated to each of the plurality of cache servers 2. In FIG. 1A, n (n=integer) content servers are provided. One cache server 3(A,B) to which content A and B are copied and (m−2) (m=integer) cache servers 3 to which content B is copied are connected to content server 2b(B) having content B. One cache server, cache server 3a(A), to which content A is copied, is connected to a content server 2a(A) having content A. Another cache server, cache server 3b, to which content A is copied, is subordinated to cache server 3a. 
For example, when the load of cache server 3a is heavy because cache server 3b is subordinated to cache server 3a, cache server 3a transfers switch request T1 to request the higher-order content server 2a to switch the connection of cache server 3b as its load source. Upon receipt of switch request T1, content server 2a transfers switch request T2 to each cache server 3 to search for the connection destination of cache server 3b. Then, when content server 2a has received response T3 from a cache server 3 which can be switched, content server 2a transfers response T4 to the cache server 3a to notify cache server 3a of the switchable cache server 3. Upon receipt of response T4, the switchable cache server 3 transfers switch request T5 to cache server 3b and re-establishes the logic link as shown in FIG. 1B.
Specifically, in FIG. 1B the total number of cache servers 3(A,B) to which content A and B are copied and cache servers 3(B) to which content B is copied becomes m−1 and a logic link for connecting cache server 3b to cache server 3(A,B) is established.
However, the method for negotiating whether a logic link can be established with a plurality of content servers having different content and re-establishing a logic link between it and the switchable cache server 3 can equalize its load, but the total load of the entire network remains unchanged, which is a problem.
Patent Document 1: Japan Patent Application No. 2003-186776
Patent Document 1: Japan Patent Application No. 2007-257357